236 BELL SYSTEM TECHNICAL JOURNAL 



I will just refer to one additional case, because it is of very recent dis- 

 covery and relates to that rare element which is so helpful in magnetics 

 and seems to be so useless for anything else: I mean gadolinium. 

 Metallic gadolinium has a value of amounting to about 300° absolute. 

 Well, last spring Trombe at Strasbourg investigated this metal at low 

 temperatures and found that it, too, is ferromagnetic, even more so 

 than iron itself. Incidentally most of the rare-earth elements have not 

 yet been prepared in pure metallic form, and it looks as though we 

 might almost count on turning up some more cases of this kind. All 

 this brings me to the question oi Jerromagnetism. 



I do not suppose that any of my readers thinks that it is ferromagne- 

 tism of which I have thus far been speaking, but for the sake of com- 

 pleteness I will give the name : up to this point we have been considering 

 paramagnetic bodies, and explaining their behavior by the orientations 

 of atoms in fields. Now we turn to the properties of iron, cobalt, 

 nickel, various alloys and compounds of these, various alloys con- 

 taining manganese, and gadolinium: the ferromagnetic substances. 



The most confusing thing about ferromagnetism — at least if my own 

 experience as a student is any guide — the most confusing thing is, that 

 the /-vs-i7 curve of a ferromagnetic substance reminds one of the sort of 

 thing that the Langevin theory is meant to explain, and yet it is not 

 that sort of thing at all. One looks at the Langevin curve with its 

 approach to saturation, and then one thinks of the curve for iron wath 

 its approach to saturation, and one cannot help but think that the two 

 must correspond to each other except for minor and trivial details. 

 Well, they do not. They differ not alone in trivial details, but in every 

 possible way, excepting the solitary common feature of the approach 

 to a horizontal asymptote. 



It is really impossible to put this statement too strongly. The 

 Langevin curve and the iron curve differ in shape, as any sketch (cf. 

 Figs. 1 and 2) will show^ They difTer utterly in scale. If I were to 

 start to put a Langevin curve on the same plot where an iron curve 

 appears with suitable detail, not only would it be sensibly linear for 

 thousands and thousands of miles, but it would not even rise appreci- 

 ably ofT the axis for hundreds of miles. Conversely if I had tried to put 

 the curve for a ferromagnetic body upon the same graph as the Lange- 

 vin curve, the former would have consisted only of the axis of ordinates 

 plus the horizontal asymptote. Finally, the temperature relations are 

 all wrong. I told you that in the Langevin curve the slope near the 

 origin varies inversely as temperature, and I left you to infer that the 

 ordinate at saturation is independent of temperature. In the curve for 

 iron, the slope near the origin goes up with the temperature, and the 



